Assessment of Uncertainties in Modelling Land Use Change with an Integrated Cellular Automata–Markov Chain Model

Uncertainty in future land use change modelling is crucial to study as it may result in varying spatial characteristics and features of the model. An integrated land use change model combines different modelling techniques and strengths. However, combined model uncertainties may significantly affect the accuracy of the model results. In this study, uncertainties resulting from spatial resolution and proportional errors of the input maps, as well as from iteration number of the cellular automata (CA) employing an integrated CA–Markov chain (CA-MC) model have been explored. The model uncertainty was quantified by comparing simulated maps to a classified land use map with the help of kappa statistics and confusion matrix. Further, correlation analysis was performed between kappa coefficients and land use simulations. The results show that the input data uncertainty (spatial resolution and proportional error) has a higher influence on land use simulation as compared to the CA-model parameter uncertainty (iteration number). The confusion matrix and the percent deviation from simulated land use map showed that variation in the major land use classes, namely agriculture (69.2 to 66.1%), dense forest (14.8 to 12%), degraded forest (12.6 to 14.5%), and barren land (1.31 to 5.96%), is required to be taken into account to reduce the model simulation error and hence the uncertainties. This study reveals that input datasets of fine spatial resolution and a low proportional error as well as to a lesser extent a high number of iterations can be recommended to minimize uncertainty in the CA-MC modelling.